goodfeli/adversarial
Reference implementation of Generative Adversarial Networks (GAN), the seminal 2014 paper by Goodfellow et al., built with Theano and Pylearn2.

Velocity · 7d
+0.9
★ / day
Trend
→steady
star history
This repository contains the original reference code and hyperparameters for the Generative Adversarial Networks paper. It includes YAML config files for training GAN models on various datasets and a script for estimating log-likelihood using Parzen density estimation. The code requires Theano and Pylearn2 dependencies and trains generative models via adversarial training between a generator and discriminator network.